Graph-Based Hub Gene Selection Technique Using Protein Interaction Information: Application to Sample Classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate disease diagnosis and prediction rely on classifying gene expression profiles.
• High-dimensional gene expression data necessitates robust feature selection for identifying critical genes.

Purpose of the Study:

• To develop an effective feature selection algorithm for sample classification using gene expression data.
• To leverage protein-protein interaction (PPI) information and graph mining for identifying significant genes.

Main Methods:

• Genes were clustered using a multi-objective optimization approach integrating gene expression and PPI data.
• Protein-protein interaction confidence scores were incorporated into the Goldberg algorithm for feature selection.
• Hub genes, identified by node degree, were utilized for sample classification with various machine learning models.

Main Results:

• The proposed method effectively identifies topologically and functionally significant hub genes.
• Classification performance was evaluated using metrics like accuracy, sensitivity, specificity, precision, F-measure, and Mathews correlation coefficient.
• Comparative analysis demonstrated the proposed approach's efficiency over existing methods.

Conclusions:

• The developed feature selection method enhances sample classification accuracy in disease prediction.
• The identified hub genes and their modules show strong biological significance, validating the approach's robustness.